Model Zoo API

[kundanapillari]

Python-API-Script

Onnx Python API Script used to download and save pretrained models from onnx model zoo. Retrieves metadata after the model is successfully downloaded.

Features

get_model_versions(MODEL_FOLDER_NAME) - Retrieves an array of all the versions of the specificed model folder in the Onnx Model Zoo.

get_pretrained() - Downloads and saves specific onnx models to desired path.

get_metadata() - Retrieves metadata of the onnx model.

Intialization and Usage

Initiate the object by calling onnx_zoo class name.

OBJECT_NAME = onnx_zoo()

The Python scipt will then ask to input the model folder name. When inputting the name, it should be in all lowercase.

Enter Model Name: resnet

After model folder name is inputted, the script will output all the model versions that exist in the folder.

['resnet101-v1-7', 'resnet101-v2-7', 'resnet152-v1-7', 'resnet152-v2-7', 'resnet18-v1-7', 'resnet18-v2-7', 'resnet34-v1-7', 'resnet34-v2-7', 'resnet50-caffe2-v1-3', 'resnet50-caffe2-v1-6', 'resnet50-caffe2-v1-7', 'resnet50-caffe2-v1-8', 'resnet50-caffe2-v1-9', 'resnet50-v1-7', 'resnet50-v2-7']

From the array of versions, input the version that will be downloaded and input the local directory that the model will be saved in.

Enter model name from options: resnet101-v1-7 
Enter saved path: /Users/name/Downloads

To download the model and output its metadata, run the following functions:

OBJECT_NAME.get_pretrained()
OBJECT_NAME.get_metadata()

Installation

Install onnx to check models

pip install onnx

Install onnxruntime to run onnx models

pip install onnxruntime

Contributors

• Kundana Pillari
• Shirley Su
• Jennifer Wang

---

[lgtm-com]

This pull request introduces 9 alerts when merging 9d0b826 into f1cb2cf - view on LGTM.com

new alerts:

• 5 for Unused import
• 3 for Except block handles 'BaseException'
• 1 for Unused argument in a formatting call

[lgtm-com]

This pull request introduces 3 alerts when merging 652225c into f1cb2cf - view on LGTM.com

new alerts:

• 3 for Except block handles 'BaseException'

[lgtm-com]

This pull request introduces 3 alerts when merging 8c06863 into adef8dd - view on LGTM.com

new alerts:

• 3 for Except block handles 'BaseException'

[lgtm-com]

This pull request introduces 3 alerts when merging dade22a into adef8dd - view on LGTM.com

new alerts:

• 3 for Except block handles 'BaseException'

[jcwchen]

Thank you for this PR. It is an awesome work.
One question:
Are these model urls here hard-coded?

If yes, I think maybe these urls can be parsed from onnx/models automatically. (finding those onnx models under model/ directory)
In that case, this API can be automatically updated while new models are introduced to model zoo.

[kundanapillari]

We used the generate_models.py file to generate the links of the models automatically. It is included in the PR.

[vinitra-zz]

@kundanapillari - is it necessary to include this dictionary directly? Since we've included the scripts to dynamically generate these files, can they be generated in the initialization step?

The problem is that we'll continually have to keep updating this dictionary of models if we check this in.

[shirleysu8]

I think one issue with having the dictionary be generated in the initialization step is that the user has to have an updated cloned repository for it to be able to traverse through the model zoo and create the dict.

[vinitra-zz]

great start!

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ kundanapillari
❌ vinitra
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[prasanthpul]

Superceded by #455